The approach requires transfer function of system and analysis in frequency domain with BODE analysis or Root Locus. Then gains can be computed and entered. If system is dynamic changing load, then a smart sweep in startup and step function test can be used to optimize stability on startup. But temperature rise in magnetics or variation of magnet strength may alter gain of PID loop for example, so this requires compensation.
The approach requires transfer function of system and analysis in frequency domain with BODE analysis or Root Locus. Then gains can be computed and entered. If system is dynamic changing load, then a smart sweep in startup and step function test can be used to optimize stability on startup. But temperature rise in magnetics or variation of magnet strength may alter gain of PID loop for example, so this requires compensation.
By monitoring the response of the system, you can determine whether there is acceptable overshoot/response time, etc. Then you can tune the system using any number of methodologies, for example: Good Gain, Zeigler-Nichols, Cohen-Coon or, my favorite, experience. You don't HAVE to know the system transfer function, and in many cases it's impossible to know.
By monitoring the response of the system, you can determine whether there is acceptable overshoot/response time, etc. Then you can tune the system using any number of methodologies, for example: Good Gain, Zeigler-Nichols, Cohen-Coon or, my favorite, experience. You don't HAVE to know the system transfer function, and in many cases it's impossible to know.
I recall Maxtor and Toshiba used variations of this to measure different overshoot and setting times on fast servo seek calibration with varying step and directions, as the system was non-linear . Some problems arose with drifting gain parameters from magnetic variations with temp so the rotary voice coil would under certain conditions, write data off track from excess overshoot, if the target sector was in sync with servo seek complete.... But only in one direction and only in one zone and ony some of the time. It took me 1day to prove it and 6 mos for the Corporation to get them to fix it.